Tests Conditional on Imbalance with Biased Coin Designs.
Abstract
Distributional properties of the treatment assignment variables T sub 1, ..., T sub n under Efron's (1971) biased coin design are derived. These properties are conditional on the terminal imbalance of the treatment allocation. Recursive procedures are presented for obtaining the conditional moments of T sub 1, ..., T sub n. Based on these results, large-sample test statistics are proposed for the randomization test of the null hypothesis of no treatment difference. In contrast to Efron's test statistic, the approximations herein proposed are applicable when there is a treatment allocation imbalance. Keywords: markov chains; chemical tests; conditional means and covariances; computerized simulation.
Document Details
- Document Type
- Technical Report
- Publication Date
- Jul 01, 1986
- Accession Number
- ADA174777
Entities
People
- Edsel Pena
- Myles Hollander
Organizations
- Florida State University